CN118279174A - Medical image processing method, device, equipment and storage medium - Google Patents

Abstract

The embodiment of the application provides a medical image processing method, a medical image processing device, medical image processing equipment and a storage medium. The method comprises the following steps: acquiring an initial medical image to be processed, and preprocessing the initial medical image to obtain a target medical image; analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer; converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result; and outputting the target medical image after noise reduction through the target diffusion model based on the first solving result. The embodiment of the application aims to process the medical image, so that the medical image noise is reduced, the medical image quality is improved, and the processing efficiency of the medical image is improved.

Description

Medical image processing method, device, equipment and storage medium

Technical Field

The present application relates to the technical field of intelligent medical science, and in particular, to a medical image processing method, a medical image processing device, a computer device, and a computer readable storage medium.

Background

Because the medical image can be polluted by different noise degrees in the processes of acquisition, transmission, display and the like, the medical image is required to be reduced in noise, so that the medical image is clearer, and the judgment capability of doctors on illness conditions is improved.

Conventional noise reduction methods often rely on a priori information or model assumptions, resulting in limited noise reduction. In recent years, diffusion models are widely applied to medical image processing, and can reduce noise of medical images without relying on prior information, however, the existing diffusion model-based medical image noise reduction efficiency is low.

Disclosure of Invention

The application provides a medical image processing method, a medical image processing device, a computer device and a computer readable storage medium, which aim to process medical images, reduce medical image noise, improve medical image quality and improve medical image processing efficiency.

To achieve the above object, the present application provides a method for processing a medical image, the method comprising:

Acquiring an initial medical image to be processed, and preprocessing the initial medical image to obtain a target medical image;

Analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer;

Converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result;

And outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

To achieve the above object, the present application also provides a medical image processing apparatus, including:

The acquisition module is used for acquiring an initial medical image to be processed and preprocessing the initial medical image to obtain a target medical image;

the analysis module is used for analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer;

The solving module is used for converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result;

And the output module is used for outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

In addition, to achieve the above object, the present application also provides a computer apparatus including a memory and a processor; the memory is used for storing a computer program; the processor is configured to execute the computer program and implement the steps of the medical image processing method according to any one of the embodiments of the present application when the computer program is executed.

In addition, to achieve the above object, the present application further provides a computer-readable storage medium storing a computer program, which when executed by a processor, causes the processor to implement the steps of the medical image processing method according to any one of the embodiments of the present application.

According to the medical image processing method, the medical image processing device, the computer equipment and the computer readable storage medium disclosed by the embodiment of the application, the initial medical image to be processed can be acquired, and then the initial medical image is preprocessed to obtain the target medical image. Further, the target medical image can be analyzed through the target diffusion model, a random differential equation in the target diffusion model is converted into a neural ordinary differential equation, and then the neural ordinary differential equation is solved based on the target medical image, so that a first solving result is obtained; thus, the target medical image after noise reduction can be obtained through the target diffusion model output based on the first solving result. According to the application, the random differential equation in the target diffusion model is converted into the neural ordinary differential equation, so that the calculated amount in the solving process can be reduced, the efficiency of the target diffusion model for outputting the target medical image after noise reduction is improved, and the processing efficiency of the medical image can be improved while the quality of the medical image is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic view of a scenario of a medical image processing method according to an embodiment of the present application;

FIG. 2 is a schematic diagram of steps of a medical image processing method according to an embodiment of the present application;

FIG. 3 is a flowchart of a medical image processing method according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a step of constructing a target diffusion model according to an embodiment of the present application;

FIG. 5 is a schematic step diagram of another medical image processing method according to an embodiment of the present application;

FIG. 6 is a schematic block diagram of a medical image processing apparatus provided by an embodiment of the present application;

Fig. 7 is a schematic block diagram of a computer device provided by an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

The flow diagrams depicted in the figures are merely illustrative and not necessarily all of the elements and operations/steps are included or performed in the order described. For example, some operations/steps may be further divided, combined, or partially combined, so that the order of actual execution may be changed according to actual situations. In addition, although the division of the functional modules is performed in the apparatus schematic, in some cases, the division of the modules may be different from that in the apparatus schematic.

The term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.

Some embodiments of the present application are described in detail below with reference to the accompanying drawings. The following embodiments and features of the embodiments may be combined with each other without conflict.

As shown in fig. 1, the medical image processing method provided by the embodiment of the application can be applied to an application environment as shown in fig. 1. The application environment includes a terminal device 110 and a server 120, where the terminal device 110 may communicate with the server 120 through a network. Specifically, the server 120 can acquire an initial medical image to be processed, and perform a preprocessing operation on the initial medical image to obtain a target medical image; analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer; further converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result; finally, based on the first solution result, outputting the target medical image after noise reduction through the target diffusion model, and sending the target medical image after noise reduction to the terminal equipment 110. The server 120 may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms. The terminal device 110 may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein.

Referring to fig. 2 and fig. 3, fig. 2 is a schematic diagram illustrating steps of a medical image processing method according to an embodiment of the present application; fig. 3 is a flowchart of a medical image processing method according to an embodiment of the present application. The medical image processing method can be applied to computer equipment, and therefore noise reduction processing is conducted on the medical image.

As shown in fig. 2, the medical image processing method includes steps S11 to S14.

Step S11: and acquiring an initial medical image to be processed, and preprocessing the initial medical image to obtain a target medical image.

The initial medical image may be a scanned image of a patient, and may be obtained by scanning with various devices such as an X-ray machine, a CT scanner, an MRI apparatus, etc., which is not limited in the present application.

Further, the target medical image is a scanned image of the patient after the preprocessing operation.

In particular, to enable better identification and processing of scanned images of a patient, pre-processing operations may be performed on scanned images of a patient. The preprocessing operation includes at least one of image cropping operation, contrast adjustment operation, gray-scale image conversion operation, and geometric transformation operation, which is not limited in the present application.

In the embodiment of the application, the scanning image of the patient to be processed can be obtained, and then the scanning image of the patient is preprocessed, so that the scanning image of the patient can be better identified and processed, and the noise reduction effect of the scanning image of the patient is improved.

Step S12: analyzing the target medical image by a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and random differential equations are arranged in the random differential equation layer.

After the target medical image is obtained, it may be input to the target diffusion model to enable the target diffusion model to analyze the target medical image.

The present application is not limited to the target diffusion model. For example, the target flooding may be FCN (Fully Convolutional Networks, full convolutional network) network, unet network, etc., and the present application is described by taking the target flooding model as Unet network as an example.

Unet network a U-shaped network structure is used in a large number of segmentation fields. Unet the network is a network constructed on the basis of the FCN, and the U-shaped structure of the network solves the defect that the FCN cannot receive the context information and the position information. The Unet network adopts the fusion operation of the low-level feature map and the rear high-level feature map, the completely symmetrical U-shaped structure enables front and rear features to be fused thoroughly, the high-resolution information and the low-resolution information are added in the target picture, the low-resolution information in the down sampling process and the high-resolution information in the up sampling process are combined, and in addition, the bottom information is filled through the fusion operation to improve the segmentation precision.

Furthermore, the Unet network further comprises a random differential equation, and the random differential equation is solved based on the target medical image, so that the noise reduction of the target medical image can be realized, and the quality of the target image is improved.

In an embodiment of the present application, the target medical image may be analyzed by a target diffusion model, wherein the target diffusion model comprises a layer of random differential equations in which random differential equations are arranged. The random differential equation is solved based on the target medical image, so that the noise reduction of the target medical image can be realized, the quality of the target image is improved, a doctor can analyze the scanned image of a patient more accurately and make diagnosis, the medical service quality is improved, the misdiagnosis rate is reduced, and better medical service is provided for the patient.

Step S13: and converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result.

The first solving result is a primary solving result. Further, the first solution result includes gray values of the target medical image.

Specifically, a random differential equation of a random differential equation layer in the Unet network can be converted into a neural ordinary differential equation, and then the neural ordinary differential equation is solved in the iteration of the target medical image, so that a first solving result comprising the gray value of the target medical image is obtained.

It should be noted that the neural ordinary differential equation is a derivative of the hidden state parameterized using a neural network. Wherein the derivative of the parameterized hidden state similarly builds a continuous hierarchy and parameters, rather than discrete hierarchies. Thus, the parameter is also a continuous space, and the gradient and update parameter do not need to be propagated in layers. That is, the neural differential equation does not store any intermediate results during the forward propagation, only requires approximately constant levels of memory cost. Therefore, the calculation amount in the solving process can be reduced by converting the random differential equation into the neural ordinary differential equation, and the processing efficiency of the target medical image is further improved.

Optionally, the stochastic differential equation comprises a stochastic term, converting the stochastic differential equation to a neural ordinary differential equation, comprising: sampling the random term in a randomizing sampling mode, and further determining target probability distribution of the sampled random term in a Gaussian distribution mode; based on the target probability distribution, the random differential equation is converted into a neural ordinary differential equation.

It should be noted that the randomized sampling is an approximate inference method based on numerical sampling. Samples of the generally uniformly distributed uniformity (0, 1), known as the rand-like function, may be generated by a linear congruence generator; while other random distributions can be obtained by some function transformation on the basis of uniform distribution, for example, normal distribution can be obtained by Box-Muller transformation. However, this transformation relies on calculating the inverse function of the integral of the target distribution, which is difficult to obtain by simple transformation when the form of the target distribution is complex or high-dimensional. Therefore, when a problem cannot be solved accurately by an analytical method, only an approximate solution of the problem can be inferred, and the randomized sampling is a powerful method for solving the approximate solution.

Further, gaussian distribution, also known as normal distribution, is the most important continuous probability distribution in statistics. In physical science and economics, the distribution of large amounts of data is typically gaussian-like, and therefore can be preferentially approximated or precisely described by a gaussian distribution when the underlying distribution pattern of the data is unclear.

In the embodiment of the application, the random term can be sampled in a randomizing sampling mode, and then the target probability distribution of the sampled random term is determined in a Gaussian distribution mode. In this manner, the random differential equation may be converted to a neural ordinary differential equation for solution over a range based on the target probability distribution.

Optionally, solving the neural ordinary differential equation based on the target medical image to obtain a first solution result includes: and solving a neural ordinary differential equation based on the target medical image by a preset method to obtain a first solving result.

It should be noted that the present application is not limited to the preset method, and includes at least one of an Euler-Maruyama method, a Milstein method, and a numerical solution method, and the present application is described by taking the preset method as an Euler-Maruyama method as an example.

The Euler-Maruyama method is a first order numerical method that is used to solve the ordinary differential equation (i.e., the initial problem) for a given initial value. It is the most basic type of phenotype method (Explicit method) that solves the numerical integration of ordinary differential equations. The basic idea is iteration, which is divided into an advancing Euler method, a retreating Euler method and a modified Euler method. The iteration is successive substitution, the required solution is finally obtained, a certain precision is achieved, and the error can be easily calculated based on an Euler-Maruyama method.

In the embodiment of the application, the neural frequent differential equation can be solved based on the target medical image based on the Euler-Maruyama method to obtain a first solving result, so as to realize the processing of the target medical image.

Step S14: and outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

After the random differential equation in the target diffusion model is improved to the neural ordinary differential equation, the neural ordinary differential equation in the target diffusion model can be solved through iteration, and a first solving result can be obtained. And then, based on the first solving result, outputting the target medical image after noise reduction through the target diffusion model. In this way, smoothing and noise reduction of the target medical image is achieved.

Optionally, on the basis of the above embodiment, the method further includes, after solving the neural ordinary differential equation based on the target medical image to obtain a first solution result: continuously solving a neural ordinary differential equation based on the gray value of the target medical image to obtain a second solving result, wherein the second solving result is in a preset result range; and outputting the first medical image through the target diffusion model based on the second solving result.

Specifically, after the first solving result including the gray value of the target medical image is obtained, the neural ordinary differential equation can be further solved based on the gray value of the target medical image, so as to obtain the second solving result. The second solution result is within a preset result range, for example, the second solution result satisfies a preset convergence condition, or the second solution result satisfies a preset iteration number. In this way, the first medical image may be obtained by the target diffusion model output based on the second solution result.

In the embodiment of the application, after the first solving result is obtained, the neural ordinary differential equation can be further solved based on the gray value of the target medical image, so that the second solving result reaches the preset iteration times or convergence conditions, and the first medical image is obtained through the output of the target diffusion model based on the second solving result. In this way, smoothing and noise reduction of the target medical image is achieved.

The medical image processing method disclosed by the embodiment of the application can acquire the initial medical image to be processed, so as to preprocess the initial medical image and obtain the target medical image. Further, the target medical image can be analyzed through the target diffusion model, a random differential equation in the target diffusion model is converted into a neural ordinary differential equation, and then the neural ordinary differential equation is solved based on the target medical image, so that a first solving result is obtained; thus, the target medical image after noise reduction can be obtained through the target diffusion model output based on the first solving result. According to the application, the random differential equation in the target diffusion model is converted into the neural ordinary differential equation, so that the calculated amount in the solving process can be reduced, the efficiency of the target diffusion model for outputting the target medical image after noise reduction is improved, and the processing efficiency of the medical image can be improved while the quality of the medical image is improved.

With continued reference to fig. 4, fig. 4 is a schematic diagram illustrating steps for constructing a target diffusion model according to an embodiment of the present application. As shown in fig. 4, the construction of the target diffusion model may be achieved through steps S21 to S24.

Step S21: an initial diffusion model is obtained.

Step S22: and carrying out iterative training on the initial diffusion model to extract data characteristics, and calculating to obtain a loss function.

Step S23: and carrying out iterative training on the loss function by using a preset method with the aim of reducing the loss function value until the expected threshold value specification is met.

The initial diffusion model, that is, the diffusion model before the iterative training, may include, for example, an FCN network, unet network, and the like.

It will be appreciated that in order to train a higher accuracy target diffusion model, the initial diffusion model may be iteratively trained in such a way that the loss function is continually reduced until the loss function meets the desired threshold specification.

The preset method and the expected threshold value are not limited, and for example, the preset method may be a gradient descent algorithm, a batch gradient descent algorithm, a random gradient descent algorithm, etc., and the gradient descent algorithm is described as an example.

The purpose of the gradient descent algorithm is to find the minimum of the loss function, or to converge to the minimum, in an iterative manner. The gradient descent algorithm geometrically, that is, where the function changes most rapidly, decreases most rapidly along the opposite direction of the vector, so that the function minimum is more easily found. Based on this, in the embodiment of the present application, repeated iterative training may be performed on the initial diffusion model by using a gradient descent algorithm, so that the loss function is continuously reduced, thereby reducing the error of the calculation result.

Step S24: and constructing a target diffusion model based on the loss function after iterative training.

After the loss function is obtained by calculation of the initial diffusion model and the loss function is trained in an iterative mode, the target diffusion model can be built based on the loss function after the iterative training.

In the embodiment of the application, the loss function is continuously reduced by repeatedly and iteratively training the initial diffusion model, and the target diffusion model is constructed based on the iteratively trained loss function, so that a target medical image with higher precision is output based on the target diffusion model, and a better noise reduction effect of the target medical image is realized.

With continued reference to fig. 5, fig. 5 is a schematic diagram illustrating steps of another medical image processing method according to an embodiment of the application. As shown in fig. 5, the noise reduction processing of the medical image can also be achieved through steps S15 to S19.

Step S15: and determining quality parameters corresponding to the target medical image after noise reduction through a preset method.

Wherein the quality parameters include, but are not limited to, signal-to-noise ratio, contrast, resolution, etc., as the application is not limited in this regard.

Furthermore, the method is not limited to the preset method, for example, the signal-to-noise ratio corresponding to the target medical image after noise reduction can be obtained by calculating the ratio of the power spectrum of the signal to the power spectrum of the noise; or obtaining the contrast of the target medical image by a method of calculating the gray value; or calculate corresponding quality parameters based on the target medical image, etc.

Step S16: judging whether the quality parameters meet preset standards or not;

because the quality parameters corresponding to the target medical image after noise reduction determine the quality and the fineness of the image, whether the quality parameters corresponding to the target medical image after noise reduction meet preset standards can be further judged to determine whether the target medical image after noise reduction needs to be processed in the next step.

It should be noted that the present application is not limited to the preset standard. For example, the preset criteria may be a signal to noise ratio of not less than 70DB, and/or a contrast ratio of greater than 2000:1, and/or a resolution of not less than 960 x 540.

Step S17: and when the quality parameters do not meet preset standards, analyzing the target medical image after noise reduction through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer.

Step S18: and converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image after noise reduction to obtain a third solving result.

Step S19: and outputting the second medical image through the target diffusion model based on the third solving result.

When the third solving result is the second noise reduction, solving the neural ordinary differential equation to obtain a calculating result; the second medical image is a medical image obtained after the secondary noise reduction.

Specifically, when the quality parameter of the target medical image after noise reduction does not meet the preset standard, the fact that noise reduction processing is needed to be performed on the target medical image after noise reduction again at the moment is indicated, so that the target medical image after noise reduction can be analyzed through the target diffusion model, a random differential equation in the target diffusion model is further converted into a neural ordinary differential equation, and the neural ordinary differential equation is solved based on the target medical image after noise reduction, and a third solving result is obtained. And finally, outputting and obtaining a second medical image through the target diffusion model based on the third solving result.

It should be noted that, for the related art description of step S17 to step S19, please refer to the above-mentioned step S12 to step S14, and the description thereof is omitted herein for avoiding repetition.

In the embodiment of the application, the quality evaluation, such as calculation of the quality parameters of signal-to-noise ratio, contrast, resolution and the like, of the medical image after noise reduction can be performed to determine whether the medical image after noise reduction meets the preset standard. If the medical image after noise reduction does not meet the preset standard, the method can be returned to carry out noise reduction on the medical image after noise reduction again so as to obtain the medical image with higher noise reduction effect.

It can be understood that the medical image processing method provided by the application can realize CT scanning image noise reduction: since CT scanning has an important role in diagnosing many diseases (e.g., lung diseases, brain diseases, etc.), CT images are susceptible to noise. Therefore, by applying the medical image processing method provided by the application, the definition of the CT image can be improved, and a doctor can be helped to diagnose the illness state more accurately.

The medical image processing method provided by the application can also realize the noise reduction of the MRI image: magnetic Resonance Imaging (MRI) is a noninvasive diagnostic technique that is widely used in the fields of nerve, musculoskeletal, cardiovascular, etc. However, MRI images are also susceptible to noise interference. By applying the medical image processing method provided by the application, the noise in the MRI image can be reduced, and the image quality can be improved, so that a more reliable diagnosis basis is provided for doctors.

The medical image processing method provided by the application can also realize the noise reduction of the X-ray image: x-ray examination is a common diagnostic method such as chest X-ray, fracture examination, etc. However, noise in the X-ray image may interfere with the physician's judgment of the lesion. By applying the medical image processing method provided by the application, the quality of the X-ray image can be improved, so that a doctor can more easily identify a focus, and the diagnosis accuracy is improved.

The medical image processing method provided by the application can also realize the noise reduction of the ultrasonic image: ultrasound examination is a non-invasive imaging technique that is widely used in obstetrical, cardiovascular, urological, and other fields. However, noise in the ultrasound image may affect the physician's judgment of the lesion. By applying the medical image processing method provided by the application, the definition of the ultrasonic image can be improved, and doctors can be helped to diagnose diseases more accurately.

Referring to fig. 6, fig. 6 is a schematic block diagram of a medical image processing apparatus according to an embodiment of the present application. The medical image processing device may be configured in a server for executing the aforementioned medical image processing method.

As shown in fig. 6, the medical image processing apparatus 200 includes: the system comprises an acquisition module 201, an analysis module 202, a solving module 203 and an output module 204.

An acquisition module 201, configured to acquire an initial medical image to be processed, and perform a preprocessing operation on the initial medical image to obtain a target medical image;

an analysis module 202, configured to analyze the target medical image through a target diffusion model, where the target diffusion model includes a random differential equation layer in which a random differential equation is set;

The solving module 203 is configured to convert the random differential equation into a neural ordinary differential equation, and solve the neural ordinary differential equation based on the target medical image, so as to obtain a first solving result;

And the output module 204 is configured to output the denoised target medical image through the target diffusion model based on the first solution result.

The solving module 203 is further configured to sample the random term by using a randomized sampling manner, and further determine a target probability distribution of the random term after sampling by using a gaussian distribution manner; the random differential equation is converted to the neural ordinary differential equation based on the target probability distribution.

The solving module 203 is further configured to solve the neural ordinary differential equation based on the target medical image by using a preset method to obtain the first solving result, where the first solving result includes a gray value of the target medical image, the preset method includes at least one of an Euler-Maruyama method, a Milstein method, and a numerical solving method, and based on the gray value of the target medical image, the solving of the neural ordinary differential equation is continued to obtain a second solving result, where the second solving result is within a preset result range.

The output module 204 is further configured to output, based on the second solution result, a first medical image through the target diffusion model.

The acquisition module 201 is further configured to acquire an initial diffusion model; performing iterative training on the initial diffusion model to extract data characteristics, and calculating to obtain a loss function; performing iterative training on the loss function by using a preset method with the aim of reducing the loss function value until the expected threshold value specification is met; and constructing the target diffusion model based on the loss function after iterative training.

The analysis module 202 is further configured to perform a feature extraction operation on the target medical image to obtain a corresponding target image feature; and analyzing the target image features through the target diffusion model.

The output module 204 is further configured to determine, by using a preset method, a quality parameter corresponding to the target medical image after noise reduction; judging whether the quality parameters meet preset standards or not; when the quality parameter does not meet a preset standard, analyzing the target medical image through the target diffusion model, and determining a second random differential equation corresponding to the target medical image; converting the second random differential equation into a corresponding second neural ordinary differential equation, and solving to obtain a third solving result; and outputting and obtaining a second medical image through the target diffusion model based on the third solving result.

It should be noted that, for convenience and brevity of description, specific working processes of the above-described apparatus and each module, unit may refer to corresponding processes in the foregoing method embodiments, which are not repeated herein.

The methods and apparatus of the present application are operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.

By way of example, the methods, apparatus described above may be implemented in the form of a computer program that is executable on a computer device as shown in fig. 7.

Referring to fig. 7, fig. 7 is a schematic diagram of a computer device according to an embodiment of the application. The computer device may be a server.

As shown in fig. 7, the computer device includes a processor, a memory, and a network interface connected by a system bus, wherein the memory may include a volatile storage medium, a non-volatile storage medium, and an internal memory.

The non-volatile storage medium may store an operating system and a computer program. The computer program comprises program instructions which, when executed, cause a processor to perform any of a number of medical image processing methods.

The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

The internal memory provides an environment for the execution of a computer program in a non-volatile storage medium that, when executed by a processor, causes the processor to perform any of a number of medical image processing methods.

The network interface is used for network communication such as transmitting assigned tasks and the like. It will be appreciated by those skilled in the art that the architecture of the computer device, which is merely a block diagram of some of the structures associated with the present application, is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or less components than those shown, or may combine some of the components, or have a different arrangement of components.

It should be appreciated that the Processor may be a central processing unit (Central Processing Unit, CPU), it may also be other general purpose processors, digital signal processors (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. Wherein the general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

Wherein in some embodiments the processor is configured to run a computer program stored in the memory to implement the steps of: acquiring an initial medical image to be processed, and preprocessing the initial medical image to obtain a target medical image; analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer; converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result; and outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

In some embodiments, the processor is further configured to sample the random term by way of randomizing sampling, and further determine a target probability distribution of the random term after sampling by way of gaussian distribution; the random differential equation is converted to the neural ordinary differential equation based on the target probability distribution.

In some embodiments, the processor is further configured to solve the neural ordinary differential equation based on the target medical image by a preset method to obtain the first solution result, where the first solution result includes a gray value of the target medical image, the preset method includes at least one of an Euler-Maruyama method, a Milstein method, and a numerical solution method, and based on the gray value of the target medical image, continue to solve the neural ordinary differential equation to obtain a second solution result, where the second solution result is within a preset result range; and outputting and obtaining a first medical image through the target diffusion model based on the second solving result.

In some embodiments, the processor is further configured to obtain an initial diffusion model; performing iterative training on the initial diffusion model to extract data characteristics, and calculating to obtain a loss function; performing iterative training on the loss function by using a preset method with the aim of reducing the loss function value until the expected threshold value specification is met; and constructing the target diffusion model based on the loss function after iterative training.

In some embodiments, the processor is further configured to perform a feature extraction operation on the target medical image to obtain a corresponding target image feature; and analyzing the target image features through the target diffusion model.

In some embodiments, the processor is further configured to determine, by a preset method, a quality parameter corresponding to the denoised target medical image; judging whether the quality parameters meet preset standards or not; when the quality parameter does not meet a preset standard, analyzing the target medical image through the target diffusion model, and determining a second random differential equation corresponding to the target medical image; converting the second random differential equation into a corresponding second neural ordinary differential equation, and solving to obtain a third solving result; and outputting and obtaining a second medical image through the target diffusion model based on the third solving result.

The embodiment of the application also provides a computer readable storage medium, and a computer program is stored on the computer readable storage medium, wherein the computer program comprises program instructions, and the program instructions realize any medical image processing method provided by the embodiment of the application when being executed.

The computer readable storage medium may be an internal storage unit of the computer device according to the foregoing embodiment, for example, a hard disk or a memory of the computer device. The computer readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SMART MEDIA CARD, SMC), a Secure Digital (SD) card, a flash memory card (FLASH CARD), or the like, which are provided on the computer device.

Further, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like.

While the application has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the application. Therefore, the protection scope of the application is subject to the protection scope of the claims.

Claims (10)

1. A method of processing a medical image, the method comprising:

Acquiring an initial medical image to be processed, and preprocessing the initial medical image to obtain a target medical image;

Analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer;

Converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result;

And outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

2. The method of claim 1, wherein the random differential equation comprises a random term, the converting the random differential equation to a neural ordinary differential equation comprising:

Sampling the random term in a randomizing sampling mode, and further determining the target probability distribution of the random term after sampling in a Gaussian distribution mode;

the random differential equation is converted to the neural ordinary differential equation based on the target probability distribution.

3. The method of claim 1, wherein solving the neural ordinary differential equation based on the target medical image results in a first solution result, comprising:

solving the neural ordinary differential equation based on the target medical image by a preset method to obtain the first solving result, wherein the first solving result comprises a gray value of the target medical image, the preset method comprises at least one of an Euler-Maruyama method, a Milstein method and a numerical solving method,

The method for solving the neural ordinary differential equation based on the target medical image further comprises the following steps after obtaining a first solving result:

continuously solving the neural ordinary differential equation based on the gray value of the target medical image to obtain a second solving result, wherein the second solving result is in a preset result range;

And outputting and obtaining a first medical image through the target diffusion model based on the second solving result.

4. The method of claim 1, wherein prior to analyzing the target medical image by a target diffusion model, comprising:

Acquiring an initial diffusion model;

Performing iterative training on the initial diffusion model to extract data characteristics, and calculating to obtain a loss function;

performing iterative training on the loss function by using a preset method with the aim of reducing the loss function value until the expected threshold value specification is met;

and constructing the target diffusion model based on the loss function after iterative training.

5. The method of claim 1, wherein the analyzing the target medical image by a target diffusion model comprises:

performing feature extraction operation on the target medical image to obtain corresponding target image features;

and analyzing the target image features through the target diffusion model.

6. The method of claim 1, wherein the preprocessing operation comprises at least one of an image cropping operation, a contrast adjustment operation, a gray map conversion operation, and a geometric transformation operation.

7. The method according to claim 1, wherein after outputting the denoised target medical image through the preset diffusion model based on the solving result, the method further comprises:

determining quality parameters corresponding to the target medical image after noise reduction through a preset method;

Judging whether the quality parameters meet preset standards or not;

When the quality parameter does not meet a preset standard, analyzing the noise-reduced target medical image through the target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer;

converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the noise-reduced target medical image to obtain a third solving result;

and outputting and obtaining a second medical image through the target diffusion model based on the third solving result.

8. A medical image processing apparatus, characterized in that the medical image processing apparatus comprises:

The acquisition module is used for acquiring an initial medical image to be processed and preprocessing the initial medical image to obtain a target medical image;

the analysis module is used for analyzing the target medical image through a target diffusion model, wherein the target diffusion model comprises a random differential equation layer, and a random differential equation is arranged in the random differential equation layer;

The solving module is used for converting the random differential equation into a neural ordinary differential equation, and solving the neural ordinary differential equation based on the target medical image to obtain a first solving result;

And the output module is used for outputting the target medical image after noise reduction through the target diffusion model based on the first solving result.

9. A computer device, comprising: a memory and a processor; wherein the memory is connected to the processor for storing a program, the processor being adapted to implement the steps of the medical image processing method according to any one of claims 1-7 by running the program stored in the memory.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, causes the processor to carry out the steps of the method of processing medical images according to any one of claims 1-7.